I
N THE WILDTYPE STUDYdescribed in our paper178, we developed an approach to behavioural analyses, inspired by the way Van Os and Kapur5presented their classification of symptoms in psychotic diseases. They distinguished five domains: psychosis (positive symptoms), volition (negative symptoms), cognition, affective dysregulation and bipolar symptoms. Similar attempts to classify symptoms into domains are being made in the Research Domain Criteria (RDoC) project39. We adapted this stategy to study murine behaviour as syndromes, which is more clinically relevant than looking at single symptoms. Consequently, we created behavioural profiles of mouse models, based on multiple behavioural tests.Data calibration and normalization allow to compare independent experiments. This is im-portant, since only a limited number of animals can be tested in a single cohort and often several cohorts are needed to test different experimental conditions. In our approach, we try to overcome this problem. It can be applied to equate data from different cohorts, handled by different ex-perimenters and even from different laboratories — provided that all cohorts contain identical calibrator groups (in our case wildtype mice in enriched environment). This way we could compare not only different housing conditions, but also mice carrying different mutations or treated with different drugs. Calibration makes it also possible to compare results expressed in different scales and units (e.g. seconds, meters). Consequently, we can include different tests that measure similar behaviours into one analysis. We could also create profiles containing not only behavioural, but also histological, electrophysiological, molecular and other kinds of data. By comparing profiles of different mouse mutants, we can evaluate their relevace to particular psychiatric diseases or their aspects. For example, in our study, social defeat (SD) induced depressive-like behaviour associated with negative symptoms, isolation rearing (IR) was more relevant for positive symptoms178and Tcf4overexpression affected specifically the cognitive domain (see page 68).
We applied multivariate statistics to merge measures of similar behaviours into higher-order categories: traits, domains superdomains and symptom classes. In this process, called dimension reduction, we reveal broader patterns of behaviour and reduce the effects of correction for multiple testing on significance thresholds. We used 15 animals per group, which is the standard n in behavioural studies on mice. Power analysis showed that this number is sufficient to detect only large effect sizes. Detecting smaller effects requires testing more animals or restricting the analyses to fewer domains, to reduce the influence multiple testing on significance threshold.
We present behavioural profiles in radar charts, which allows to plot huge data sets in a single figure. By overlaying them, we can also compare experimental groups to each other. Plots can be generated at every level of dimension reduction. On the trait level they show all performed tests and are useful mainly for behavioural scientists. Plots on the domain level are more interesting for clinicians, who may compare different mouse mutants to choose the one that reflects a particular class of symptoms best.
5.1.1 Comparison of IR and SD as models of psychotic diseases
Both IR and SD induced remarkable impairments in curiosity, fear conditioning and motivation, but with greater impact of SD. Stronger overall impairment in SD mice is reflected by the higher severity score. However, some behaviours impaired by IR are unaffected or barely changed upon SD. IR mice displayed typical for isolated rodents136,139,228–230hyperactivity in OF, which may be relevant for positive symptoms of schizophrenia231. Similarly, striking hypoalgesia, repeatedly reported in isolated rodents134,135,230, is an endophenotype of schizophrenia205–208,232. The differences in behavioural profiles of IR and SD show that these two paradigms should be used to model different aspects of psychiatric diseases. While SD appears as relevant for negative symptoms, IR seems to be more suitable for positive symptoms.
It has to be noted, that our study focused on behavioural phenotype based on limited num-ber of tests. For a broader symptom coverage and higher clinical relevance, the analysis should include more behavioural tests and be supplemented with other data, e.g. EEG recordings, electro-physiology, histology or gene expression.
5.1.2 G×E-dependent cognitive deficits in Tcf4tg mice
While environmental conditions (IR, SD or EE) strongly affected several murine behaviours, effects ofTcf4overexpression were mild and restricted to cognition. Both young and agedTcf4tg mice displayed impaired fear conditioning and reversal learning upon IR, whereas EE rescued the phenotype. This influence of environment on manifestation of theTcf4-dependent deficits proves the Gene×Environment interaction in theTcf4tg mice.
AgedTcf4tg mice, in comparison to young mice, showed no additional impairments upon IR except from subtly reduced rearing and increased swim speed. Reversal learning deficits were milder in the aged cohort, but were confirmed in the delayed matching to place (DMP) test. As no clear worsening of the symptoms was apparent in 12 months oldTcf4tg mice, we conclude that the Tcf4phenotype is independent of ageing.
Tcf4tg mice upon IR and SD, displayed impaired reversal learning in MWM, confirmed by disrupted delayed matching to place. Decline in reversal learning is a measure of behavioural rigidity, or perseveration — a psychological term describingoverall, perseverance in doing some-thing to an awesome level or past an adequate point; (...) improper repeating of actions which are frequently correlated with injury to the brain’s frontal lobe, incapacity (...) to switch from one
method or process to another one154. Perseveration and reversal learning deficits are associated with dysfunction of the orbitofrontal cortex (OFC) and ventral striatum18,233–235 and have been reported in schizophrenia and psychotic disorders17,18. In mice, reversal learning is specifically disrupted bysocialdeprivation145,236.
Comparison of behavioural profiles of Tcf4tg mice subjected to IR and SD1 revealed that both IR and SD trigger cognitive impairments in theTcf4tg mice. Upon IR,Tcf4tg mice showed impaired fear memory and typical for IR behavioural rigidity145,236. Upon SD they displayed milder rigidity and pronounced impairments of spatial learning and memory recall in MWM. In the control condition (individual housing with daily handling), used as reference for SD,Tcf4tg mice displayed similar deficits to theTcf4tg IR mice (Fig. 4.15), but less severe. This is consistent with the observations in rats, that handling can diminish the effects of isolation230. It suggests thatTcf4 overexpression increases vulnerability to harsh environment and the type environmental treatment determines which brain structures, and consequently, which behaviours will be affected the most.
5.1.3 Cognitive deficits and Tcf4 expression in Tcf4C knockout mice
We generated the heterozygousTcf4knockout mouse lineTcf4C. Tcf4exons 5–6, located directly downstream the deleted exon 4, had reduced expression in these mice and the levels subsequent exons were gradually increasing. Expression patterns ofTcf4exons in wt mice differed between PFC and hippocampus: exons 5–6 were highest expressed in hippocampus, but not in PFC. Exons 5–6 showed strongest downregulation inTcf4C mice, which is in line with the exclusively hippocampal phenotype observed in the Morris water maze.
SeveralTCF4isoforms of various lengths and exon composition were found in humans45,237 and in mice200(Fig. 5.1A). Genetic analyses show that the Pitt-Hopkins syndrome is caused by deletions and nonsense mutations that occur in exons 8–2045,80,237(Fig. 5.1B). Transcription of this region is barely affected by the knockout, thus theTcf4C mice should not be considered as a PTHS model, but rather as a tool to study functions of particularTcf4isoforms.
Tcf4C animals showed no significant alterations in any of the analysed facial and body dimen-sions, even though craniofacial abnormalities and decreased body length were reported by the Sanger Institute inTcf4E homozygotic females199. However, our method — manual measurements with a calliper – may be insufficient to detect subtle alterations in the facial features. Perhaps analysis of skull landmarks238 or with the use of computer tomography239 would reveal subtle dismorphisms in theTcf4C mice.
5.1.4 Tcf4, G×E and behavioural profiling — conclusions
Behavioural experiments withTcf4overexpressing and knockout mice revealed cognitive deficits in both cases — mild and environment-dependent inTcf4tg mice; and strong, but restricted to spatial learning inTcf4C knockouts. It appears that cognition in mice depends onTcf4gene dosage in the
1theTcf4tg SD–ctrl data set was published in the MSc thesis of Ananya Chowdhury203
Figure 5.1: Comparison of PTHS mutations with the mutation inTcf4C mice. A)Figure adapted from Ensembl Genome Browser200. In the Tcf4C line the exon 4 is deleted, which leads to frame shift and formation of non-functional Tcf4 transcripts. Mouse Tcf4has numerous isoform, some of which start downstream to exon 4, e.g. short isoformsTcf4-007andTcf4-008located in the 3’ region of the gene. These isoforms may be still functional in our knockout line.B)The PTHS-associated mutations are deletions and nonsense mutations in the 3’ region of the humanTCF445,80,237. Therefore deleting exon 4 in the 5’ region of mouseTcf4may be not reflecting the situation in PTHS patients; however, is useful for studying functions of the long isoforms, e.g.Tcf4-001andTcf4-003.
bell-shape fashion46. Disrupted balance, either up- or downregulation, has negative consequences (Fig. 5.2). Similar pattern has been observed for another schizophrenia risk element240, the NRG1-ERBB4 signalling241.
It should be noted thatTcf4tg mice have limited clinical relevance. TCF4 expression may be elevated in psychotic patients73–75 and the Tcf4tg mice show some schizophrenia-relevant symptoms. However, they overexpress an intron-lessTcf4open reading frame (overexpression of the full gene is impossible due to the huge gene size), while the schizophrenia-risk SNPs in TCF4are located in the introns. Therefore theTcf4transgenic mice should be considered not as a clear-cut model of schizophrenia, but rather as a model for studyingTcf4functions in the forebrain.
Similarly,Tcf4C knockouts are not a model of the Pitt-Hopkins syndrome, but rather a tool for studying certainTcf4isoforms. Nevertheless, both mouse strains are useful for investigating the Tcf4functions in the brain and may help understand both diseases.
Figure 5.2: Bell-shaped relationship betweenTcf4expression and cognition. Tcf4dosage influences cognitive performance — either reduced or increased levels ofTcf4have detrimental effect (marked in red).